Background

• Author: <郭伊軒>
• Created At: <2025-10-30>
• Research Motivation and Context (why are we interested in the findings?)：
  本研究透過時間序列視覺化的方式，探討遊戲在不同事件（如季節促銷、DLC 發行、續作推出）下或其玩家數、追蹤者數與好評率變化對折扣時間點的影響。藉由同時觀察季節促銷與非季節折扣期間的背景變化，研究可進一步辨識不同類型促銷活動對商家折扣行為的差異性。

• Main Findings and Takeaways：

  • 在非季節性折扣時點前後，其玩家數、追蹤者數與好評率變化有折扣前下降折扣後上升的趨勢。

  • 是否發行DLC相較於是否發行續作有較高的打折機率。

• Future Direciton：

  • 比較競爭對手活動 / 市場整體狀況：同類型遊戲中打折比例、同時期平均價格下降幅度。

  • 比較不同遊戲類型或發行商之間的差異：分析不同遊戲類型或發行商在折扣策略上的差異。

  • 結合機器學習模型進行折扣預測：例如利用 Random Forest 或 XGBoost 模型預測折扣時點，以驗證變數重要性並提升實務應用價值。

# Load packages here
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from datetime import timedelta
from sklearn.preprocessing import StandardScaler

input_data_file = "/Users/10610/Desktop/114-1 資料/steam-project/discount-timing-DE.csv"
df = pd.read_csv(input_data_file)
df['Date'] = pd.to_datetime(df['Date'])
game_id = df['GameID'].unique()   # 取得所有獨特的 GameID
game_dfs = []                     # 用來裝每個遊戲的 DataFrame

for gid in game_id:
    sub_df = df[df['GameID'] == gid]  # 篩選出該 GameID 的資料
    game_dfs.append(sub_df)

Sample

df.head()

	Date	GameID	Type	MultiPlayer	Publisher	ConstantDiscount	DiscountOrNot	DiscountDuration	DiscountFreq3M	Age	...	FollowersGrowthRate1M	PositiveRateGrowthRate1W	PositiveRateGrowthRate2W	PositiveRateGrowthRate1M	DLC_sum_1W	DLC_sum_2W	DLC_sum_1M	Sequel_sum_1W	Sequel_sum_2W	Sequel_sum_1M
0	2023-05-01	10	Action	1	Valve	0	0	0	1	22.509589	...	0.003889	0.000012	-0.000014	1.098178e-05	0	0	0	0	0	0
1	2023-05-02	10	Action	1	Valve	0	0	0	1	22.512329	...	0.003913	-0.000010	-0.000036	-4.698912e-07	0	0	0	0	0	0
2	2023-05-03	10	Action	1	Valve	0	0	0	1	22.515068	...	0.003979	-0.000011	-0.000041	-8.082766e-07	0	0	0	0	0	0
3	2023-05-04	10	Action	1	Valve	0	0	0	1	22.517808	...	0.004101	-0.000012	-0.000050	-2.450820e-05	0	0	0	0	0	0
4	2023-05-05	10	Action	1	Valve	0	0	0	1	22.520548	...	0.003912	-0.000023	-0.000053	-3.754777e-05	0	0	0	0	0	0

5 rows × 29 columns

df.describe().T

	count	mean	std	min	25%	50%	75%	max
GameID	23938.0	461376.742000	298559.181056	10.000000	244850.000000	431730.000000	644930.000000	1.145360e+06
MultiPlayer	23938.0	0.464241	0.498730	0.000000	0.000000	0.000000	1.000000	1.000000e+00
ConstantDiscount	23938.0	0.214387	0.410405	0.000000	0.000000	0.000000	0.000000	1.000000e+00
DiscountOrNot	23938.0	0.019885	0.139607	0.000000	0.000000	0.000000	0.000000	1.000000e+00
DiscountDuration	23938.0	0.221196	1.715483	0.000000	0.000000	0.000000	0.000000	3.200000e+01
DiscountFreq3M	23938.0	1.797644	1.043279	0.000000	1.000000	2.000000	3.000000	6.000000e+00
Age	23938.0	7.634427	4.458471	2.389041	4.951370	6.323288	8.479452	2.484658e+01
AccumulatedPositiveRate	23938.0	0.928061	0.064186	0.738751	0.905517	0.953165	0.972651	9.929734e-01
SalePeriod	23938.0	0.146420	0.353534	0.000000	0.000000	0.000000	0.000000	1.000000e+00
DiscountDuringSale	23938.0	0.008647	0.092590	0.000000	0.000000	0.000000	0.000000	1.000000e+00
DiscountOutOfSale	23938.0	0.011237	0.105411	0.000000	0.000000	0.000000	0.000000	1.000000e+00
PlayerGrowthRate1W	23938.0	0.020047	0.288771	-0.592919	-0.063356	-0.014085	0.035173	1.131120e+01
PlayerGrowthRate2W	23938.0	0.032978	0.370759	-0.726683	-0.088489	-0.013852	0.060643	1.088149e+01
PlayerGrowthRate1M	23938.0	0.039727	0.395094	-0.768049	-0.108811	-0.009196	0.090659	7.285318e+00
FollowersGrowthRate1W	23938.0	0.001576	0.001482	-0.000137	0.000640	0.001111	0.002078	2.296777e-02
FollowersGrowthRate2W	23938.0	0.003159	0.002758	-0.000176	0.001343	0.002289	0.004157	3.623771e-02
FollowersGrowthRate1M	23938.0	0.006816	0.005480	0.000085	0.003142	0.005071	0.009104	5.282514e-02
PositiveRateGrowthRate1W	23938.0	0.000017	0.000332	-0.015190	-0.000025	0.000005	0.000044	6.059725e-03
PositiveRateGrowthRate2W	23938.0	0.000033	0.000569	-0.016712	-0.000041	0.000009	0.000077	9.389808e-03
PositiveRateGrowthRate1M	23938.0	0.000070	0.001033	-0.017811	-0.000078	0.000015	0.000151	1.392160e-02
DLC_sum_1W	23938.0	0.004679	0.071822	0.000000	0.000000	0.000000	0.000000	2.000000e+00
DLC_sum_2W	23938.0	0.009358	0.101767	0.000000	0.000000	0.000000	0.000000	2.000000e+00
DLC_sum_1M	23938.0	0.020553	0.150182	0.000000	0.000000	0.000000	0.000000	2.000000e+00
Sequel_sum_1W	23938.0	0.001170	0.034181	0.000000	0.000000	0.000000	0.000000	1.000000e+00
Sequel_sum_2W	23938.0	0.002339	0.048312	0.000000	0.000000	0.000000	0.000000	1.000000e+00
Sequel_sum_1M	23938.0	0.005013	0.070626	0.000000	0.000000	0.000000	0.000000	1.000000e+00

The actual analysis starts below

Make the graphs, summary statistics, regression model below. Make sure you have followed the guidelines as specified in 專案資料夾結構、檔案命名與文件規範.

1W

時間序列圖

for index, sub_df in enumerate(game_dfs):
    dlc_first_day = sub_df[
        (sub_df['DLC_sum_1W'] > 0) &
        (sub_df['DLC_sum_1W'].shift(1) == sub_df['DLC_sum_1W'] - 1)
    ]

    sequel_first_day = sub_df[
        (sub_df['Sequel_sum_1W'] > 0) &
        (sub_df['Sequel_sum_1W'].shift(1) == sub_df['Sequel_sum_1W'] - 1)
    ]
    fig, axes = plt.subplots(3, 1, figsize=(15, 9), sharex=True)

    # ---- 大圖標題 & 全圖Y軸標籤 ----
    gid = sub_df['GameID'].iloc[0]  # 假設每個sub_df都有GameID欄位
    fig.suptitle(f'Game {gid} Time Series', fontsize=14, y=0.95)  # 給整張圖標題
    fig.text(0.04, 0.5, 'Rate', va='center', rotation='vertical', fontsize=12)  # 給整張圖的y軸標籤

    # ---- 各子圖 ----
    axes[0].plot(sub_df['Date'], sub_df['PlayerGrowthRate1W'], label='Players', color='blue')
    axes[0].set_title('PlayerGrowthRate1W')
    axes[0].scatter(dlc_first_day['Date'],
                    dlc_first_day['PlayerGrowthRate1W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[0].scatter(sequel_first_day['Date'],
                    sequel_first_day['PlayerGrowthRate1W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[1].plot(sub_df['Date'], sub_df['FollowersGrowthRate1W'], label='Followers', color='orange')
    axes[1].set_title('FollowersGrowthRate1W')
    axes[1].scatter(dlc_first_day['Date'],
                    dlc_first_day['FollowersGrowthRate1W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[1].scatter(sequel_first_day['Date'],
                    sequel_first_day['FollowersGrowthRate1W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[2].plot(sub_df['Date'], sub_df['PositiveRateGrowthRate1W'], label='Positive Rate', color='red')
    axes[2].set_title('PositiveRateGrowthRate1W')
    axes[2].scatter(dlc_first_day['Date'],
                    dlc_first_day['PositiveRateGrowthRate1W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[2].scatter(sequel_first_day['Date'],
                    sequel_first_day['PositiveRateGrowthRate1W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)


    # ---- 標記打折期間 ----
    sale_labels = {'Seasonal Sale': False, 'Non-Seasonal Sale': False}

    for _, row in sub_df.iterrows():
        if row['DiscountOrNot'] == 1:
            start_date = row['Date']
            end_date = row['Date'] + timedelta(days=row['DiscountDuration'])

            # 判斷季節/非季節折扣
            if row['SalePeriod'] == 1:
                color = 'red'
                label = 'Seasonal Sale'
            else:
                color = 'blue'
                label = 'Non-Seasonal Sale'

            # 只讓第一次出現的區間加上 label
            for ax in axes:
                ax.axvspan(start_date, end_date, color=color, alpha=0.15,
                        label=label if not sale_labels[label] else None)

            sale_labels[label] = True


        

    # ---- 共用設定 ----
    for ax in axes:
        ax.legend()
        ax.grid(True, linestyle='--', alpha=0.3)

    plt.xlabel('Date')
    plt.xticks(rotation=45)
    plt.tight_layout(rect=[0.05, 0.05, 1, 0.95])  # 保留空間給標題和y標籤
    plt.show()

    if index == 5:
        break
    

打折的前後趨勢對比

window = 7
event_data = []

for _, row in df[df['DiscountOrNot'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1W', 'FollowersGrowthRate1W', 'PositiveRateGrowthRate1W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountDuringSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1W', 'FollowersGrowthRate1W', 'PositiveRateGrowthRate1W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountOutOfSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1W', 'FollowersGrowthRate1W', 'PositiveRateGrowthRate1W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Non-seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

2W

時間序列圖

for index, sub_df in enumerate(game_dfs):
    dlc_first_day = sub_df[
        (sub_df['DLC_sum_1W'] == 1) &
        (sub_df['DLC_sum_1W'].shift(1) != 1)
    ]

    sequel_first_day = sub_df[
        (sub_df['Sequel_sum_1W'] == 1) &
        (sub_df['Sequel_sum_1W'].shift(1) != 1)
    ]
    fig, axes = plt.subplots(3, 1, figsize=(15, 9), sharex=True)

    # ---- 大圖標題 & 全圖Y軸標籤 ----
    gid = sub_df['GameID'].iloc[0]  # 假設每個sub_df都有GameID欄位
    fig.suptitle(f'Game {gid} Time Series', fontsize=14, y=0.95)  # 給整張圖標題
    fig.text(0.04, 0.5, 'Rate', va='center', rotation='vertical', fontsize=12)  # 給整張圖的y軸標籤

    # ---- 各子圖 ----
    axes[0].plot(sub_df['Date'], sub_df['PlayerGrowthRate2W'], label='Players', color='blue')
    axes[0].set_title('PlayerGrowthRate2W')
    axes[0].scatter(dlc_first_day['Date'],
                    dlc_first_day['PlayerGrowthRate2W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[0].scatter(sequel_first_day['Date'],
                    sequel_first_day['PlayerGrowthRate2W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[1].plot(sub_df['Date'], sub_df['FollowersGrowthRate2W'], label='Followers', color='orange')
    axes[1].set_title('FollowersGrowthRate2W')
    axes[1].scatter(dlc_first_day['Date'],
                    dlc_first_day['FollowersGrowthRate2W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[1].scatter(sequel_first_day['Date'],
                    sequel_first_day['FollowersGrowthRate2W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[2].plot(sub_df['Date'], sub_df['PositiveRateGrowthRate2W'], label='Positive Rate', color='red')
    axes[2].set_title('PositiveRateGrowthRate2W')
    axes[2].scatter(dlc_first_day['Date'],
                    dlc_first_day['PositiveRateGrowthRate2W'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[2].scatter(sequel_first_day['Date'],
                    sequel_first_day['PositiveRateGrowthRate2W'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)


    # ---- 標記打折期間 ----
    sale_labels = {'Seasonal Sale': False, 'Non-Seasonal Sale': False}

    for _, row in sub_df.iterrows():
        if row['DiscountOrNot'] == 1:
            start_date = row['Date']
            end_date = row['Date'] + timedelta(days=row['DiscountDuration'])

            # 判斷季節/非季節折扣
            if row['SalePeriod'] == 1:
                color = 'red'
                label = 'Seasonal Sale'
            else:
                color = 'blue'
                label = 'Non-Seasonal Sale'

            # 只讓第一次出現的區間加上 label
            for ax in axes:
                ax.axvspan(start_date, end_date, color=color, alpha=0.15,
                        label=label if not sale_labels[label] else None)

            sale_labels[label] = True


        

    # ---- 共用設定 ----
    for ax in axes:
        ax.legend()
        ax.grid(True, linestyle='--', alpha=0.3)

    plt.xlabel('Date')
    plt.xticks(rotation=45)
    plt.tight_layout(rect=[0.05, 0.05, 1, 0.95])  # 保留空間給標題和y標籤
    plt.show()
    if index == 5:
        break

打折的前後趨勢對比

window = 7
event_data = []

for _, row in df[df['DiscountOrNot'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate2W', 'FollowersGrowthRate2W', 'PositiveRateGrowthRate2W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountDuringSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate2W', 'FollowersGrowthRate2W', 'PositiveRateGrowthRate2W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountOutOfSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate2W', 'FollowersGrowthRate2W', 'PositiveRateGrowthRate2W']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Non-seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

1M

時間序列圖

for index, sub_df in enumerate(game_dfs):
    dlc_first_day = sub_df[
        (sub_df['DLC_sum_1W'] == 1) &
        (sub_df['DLC_sum_1W'].shift(1) != 1)
    ]

    sequel_first_day = sub_df[
        (sub_df['Sequel_sum_1W'] == 1) &
        (sub_df['Sequel_sum_1W'].shift(1) != 1)
    ]
    fig, axes = plt.subplots(3, 1, figsize=(15, 9), sharex=True)

    # ---- 大圖標題 & 全圖Y軸標籤 ----
    gid = sub_df['GameID'].iloc[0]  # 假設每個sub_df都有GameID欄位
    fig.suptitle(f'Game {gid} Time Series', fontsize=14, y=0.95)  # 給整張圖標題
    fig.text(0.04, 0.5, 'Rate', va='center', rotation='vertical', fontsize=12)  # 給整張圖的y軸標籤

    # ---- 各子圖 ----
    axes[0].plot(sub_df['Date'], sub_df['PlayerGrowthRate1M'], label='Players', color='blue')
    axes[0].set_title('PlayerGrowthRate1M')
    axes[0].scatter(dlc_first_day['Date'],
                    dlc_first_day['PlayerGrowthRate1M'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[0].scatter(sequel_first_day['Date'],
                    sequel_first_day['PlayerGrowthRate1M'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[1].plot(sub_df['Date'], sub_df['FollowersGrowthRate1M'], label='Followers', color='orange')
    axes[1].set_title('FollowersGrowthRate1M')
    axes[1].scatter(dlc_first_day['Date'],
                    dlc_first_day['FollowersGrowthRate1M'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[1].scatter(sequel_first_day['Date'],
                    sequel_first_day['FollowersGrowthRate1M'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)

    axes[2].plot(sub_df['Date'], sub_df['PositiveRateGrowthRate1M'], label='Positive Rate', color='red')
    axes[2].set_title('PositiveRateGrowthRate1M')
    axes[2].scatter(dlc_first_day['Date'],
                    dlc_first_day['PositiveRateGrowthRate1M'],
                    color='gray', s=40, marker='o', label='DLC Release', zorder=5)
    axes[2].scatter(sequel_first_day['Date'],
                    sequel_first_day['PositiveRateGrowthRate1M'],
                    color='purple', s=40, marker='o', label='Sequel Release', zorder=5)
    


    # ---- 標記打折期間 ----
    sale_labels = {'Seasonal Sale': False, 'Non-Seasonal Sale': False}

    for _, row in sub_df.iterrows():
        if row['DiscountOrNot'] == 1:
            start_date = row['Date']
            end_date = row['Date'] + timedelta(days=row['DiscountDuration'])

            # 判斷季節/非季節折扣
            if row['SalePeriod'] == 1:
                color = 'red'
                label = 'Seasonal Sale'
            else:
                color = 'blue'
                label = 'Non-Seasonal Sale'

            # 只讓第一次出現的區間加上 label
            for ax in axes:
                ax.axvspan(start_date, end_date, color=color, alpha=0.15,
                        label=label if not sale_labels[label] else None)

            sale_labels[label] = True


        

    # ---- 共用設定 ----
    for ax in axes:
        ax.legend()
        ax.grid(True, linestyle='--', alpha=0.3)

    plt.xlabel('Date')
    plt.xticks(rotation=45)
    plt.tight_layout(rect=[0.05, 0.05, 1, 0.95])  # 保留空間給標題和y標籤
    plt.show()

    if index == 5:
        break
    

打折的前後趨勢對比

window = 7
event_data = []

for _, row in df[df['DiscountOrNot'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1M', 'FollowersGrowthRate1M', 'PositiveRateGrowthRate1M']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountDuringSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1M', 'FollowersGrowthRate1M', 'PositiveRateGrowthRate1M']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

window = 7
event_data = []

for _, row in df[df['DiscountOutOfSale'] == 1].iterrows():
    date = row['Date']
    event_window = df[(df['Date'] >= date - pd.Timedelta(days=window)) &
                      (df['Date'] <= date + pd.Timedelta(days=window))].copy()
    event_window['DaysFromEvent'] = (event_window['Date'] - date).dt.days
    event_data.append(event_window)

event_df = pd.concat(event_data)
avg_change = event_df.groupby('DaysFromEvent')[['PlayerGrowthRate1M', 'FollowersGrowthRate1M', 'PositiveRateGrowthRate1M']].mean()

scaler = StandardScaler()
avg_change_scaled = avg_change.copy()
avg_change_scaled[:] = scaler.fit_transform(avg_change)

avg_change_scaled.plot(title='Standardized Change Around Non-seasonal Discount Start', figsize=(10,5))
plt.axvline(0, color='red', linestyle='--', label='Discount Start')
plt.legend()
plt.show()

其他事件對折扣影響

heatmap_data = df.groupby(['DLC_sum_1W', 'Sequel_sum_1W'])['DiscountOrNot'].mean().unstack()
sns.heatmap(heatmap_data, cmap='coolwarm', annot=False)
plt.title('Discount Probability by DLC and Sequel release in 1 week')
plt.xlabel('the number of Sequel release')
plt.ylabel('the number of DLC release')
plt.show()

這些遊戲在我們指定的時間範圍內沒有同時發行DLC和續作，發行DLC打折的比率高於發出續作。

len(df[df['DLC_sum_1W'] == 2])

6

heatmap_data = df.groupby(['DLC_sum_2W', 'Sequel_sum_2W'])['DiscountOrNot'].mean().unstack()
sns.heatmap(heatmap_data, cmap='coolwarm', annot=False)
plt.title('Discount Probability by DLC and Sequel release in 2 week')
plt.xlabel('the number of Sequel release')
plt.ylabel('the number of DLC release')
plt.show()

heatmap_data = df.groupby(['DLC_sum_1M', 'Sequel_sum_1M'])['DiscountOrNot'].mean().unstack()
sns.heatmap(heatmap_data, cmap='coolwarm', annot=False)
plt.title('Discount Probability by DLC and Sequel release in 1 month')
plt.xlabel('the number of Sequel release')
plt.ylabel('the number of DLC release')
plt.show()

heatmap_data = df.groupby(['MultiPlayer', 'DiscountFreq3M'])['DiscountOrNot'].mean().unstack()
sns.heatmap(heatmap_data, cmap='coolwarm', annot=False)
plt.title('Discount Probability by MultiPlayer and Discount Frequency in 3 Month')
plt.xlabel('Discount Frequency (3 Months)')
plt.ylabel('MultiPlayer (0 = Single, 1 = Multi)')
plt.show()